Primer · vocabulary before verdicts

What is noise cancellation? Passive, active and AI-based, explained plainly

Noise cancellation is any technique — physical or computational — that removes unwanted sound from what you hear or what you transmit. The term hides three different technologies, and buying the wrong one is this category’s most common mistake.

The three technologies wearing one name

Passive isolation is physics without electronics: sealed earcups, foam tips, mass and distance. It works on everything equally and improves nothing about what your microphone sends.

Active noise cancellation is the headphone trick the ads mean. Microphones on the earcup sample the ambient drone; the driver plays that waveform inverted; the two cancel. ANC is superb against steady, low-frequency sound — engines, air conditioning — and mediocre against sudden or speech-like noise, which arrives faster than the electronics can invert it.

AI-based software cancellation is the newest arrival and the only one this site scores. A model trained on thousands of hours of speech learns the statistical shape of a human voice, then rebuilds your microphone signal with everything else subtracted — keyboard, dog, espresso machine. It runs as a virtual microphone (Krisp, NVIDIA Broadcast, SteelSeries Sonar) or as an after-the-fact repair pass on recordings (Adobe Enhance, Cleanvoice).

What is active noise cancellation actually good for?

Concentration and travel. ANC subtracts the world from your ears: the aircraft drone, the open-plan hum. What it cannot do — structurally, not as a quality failure — is make you sound better on a call, because your microphone never wears your headphones. The commuter’s mistake is buying $300 earbuds to fix a complaint their colleagues made about the café noise behind their voice.

Which direction is your problem?

One question sorts the whole category: who is suffering? If you cannot hear, that is ANC hardware or passive isolation. If others cannot hear you, that is microphone-side software — our ranked standings compare the six tools we score. If the damage is in a file that already exists, that is repair software, led by Adobe’s free Enhance tier. And if you want the live subset only — filtering the call as it happens — the noise suppression rankings are that shelf.

Why software took over the meetings problem

Remote work moved the noise problem from ears to microphones, and physics does not scale down a renovation next door. Voice-isolation models do: they run on a laptop CPU (Krisp) or a GPU’s spare tensor cores (NVIDIA Broadcast), and they improved fast because the training data — speech versus everything else — is abundant. The result, which would have sounded like marketing a decade ago, is that a free software tier now outperforms most hardware at the transmit-side problem.

Questions we actually get

What is noise cancellation in simple terms?

Any technique that removes unwanted sound from what you hear or transmit. Hardware does it with physics — sealed earcups or inverted sound waves — while software does it with signal processing that learns what a voice is and subtracts the rest.

What is active noise cancellation (ANC)?

A hardware technique: microphones sample ambient noise and speakers play its inverted waveform, cancelling it before it reaches your ear. It lives in headphones and earbuds, and it protects only your ears — not what your microphone sends.

Can software replace ANC headphones?

They solve opposite directions. ANC quiets what you hear; software like Krisp or NVIDIA Broadcast quiets what you transmit and can also clean incoming audio. Many remote workers end up with one of each.